I-Corps: Personalized recommender systems based on semantical clustering

这个I-Corps项目的更广泛的影响/商业潜力是建立一个推荐系统，该系统为文物分配价值，并为不受监管的市场中的利益相关者提供更好的信息，以降低风险和提高信任。互联网和技术通过创建数字市场并为市场监管引入更多问题，彻底改变了包括艺术行业在内的每个商业部门之间的互动性质。建议的推荐系统可用于预测投资价值，在法律的诉讼、保险案件和银行作为抵押品收费中提供商品的客观价值。 这通过更好的决策支持市场的金融稳定，并提高利益相关者对其投资的信任。它还有助于保持对工作的商业方面的控制。例如，博物馆的采购部门可以了解他们购买的艺术品的真实的价值。I-Corps项目的重点是构建一个推荐系统，利用现有的数据集和深度知识为艺术品分配价值。艺术品市场经常被错误估值，这使得建立推荐系统的经典方法由于不可靠的数据集而变得相当困难。在这些数据集上训练的分类器通常会分配夸大的价格。这里使用的人群的方法，预计会得到更可靠的结果。这类问题的个性化可以通过分配语义距离来实现。 该项目通过从挖掘用户给出的评论（例如艺术家传记和绘画图像）中添加语义特征来扩展经典特征组。对于每个语义相似的艺术家的集群，个性化推荐系统的艺术评价。群众的智慧将被用来分配一个特定的价值，以艺术作品。

The broader impact/commercial potential of this I-Corps project is to build a recommender system which assigns values to artifacts and will also provide stakeholders in unregulated markets with better information for reducing risk and improving trust. The Internet and technology have entirely changed the nature of interactions between every business sector, including the art industry, by creating a digital marketplace and introducing additional issues for markets regulation. The proposed recommender system can be used to predict value of investment, to provide an objective value of goods in legal proceedings, insurance cases, and banks as a collateral toll. This supports financial stability of markets through better decision making and improves the trust of stakeholders in their investments. It also helps to maintain control of the commercial aspects of work. For example, the acquisitions departments of museums would be served by knowing a real value for the artifacts they buy.This I-Corps project is focused on building a recommender system for assigning values to artworks using existing datasets and deep knowledge. Art markets are often misvalued which makes the classical approach to building recommender systems rather difficult due to unreliable datasets. Classifiers trained on these datasets often assign inflated prices. A crowd approach is used here which is expected to give more reliable results. Personalization for this class of problems can be achieved by assigning semantic distance. The project extends the group of classical features by adding semantic features from mining comments given by users, and, for example, from artist biographies, and images of paintings. For each cluster of semantically similar artists, a personalized recommender system for art evaluation is built. Wisdom of the crowd will be used to assign a specific value to an art piece.